Walking the Talk: How to Identify Anti-Pluralist Parties

I N S T I T U T E

Walking the Talk: How to Identify Anti-Pluralist Parties

Anna Lührmann Juraj Medzihorsky Staffan I. Lindberg

Working Paper

SERIES 2021:116

March 2021

Varieties of Democracy (V-Dem) is a new approach to conceptualization and measurement of democracy. The headquarters – the V-Dem Institute – is based at the University of Gothenburg with 23 staff. The project includes a worldwide team with 5 Principal Investigators, 19 Project Managers, 33 Regional Managers, 134 Country Coordinators, Research Assistants, and 3,500 Country Experts. The V-Dem project is one of the largest ever social science research-oriented data collection programs.

Please address comments and/or queries for information to:

V-Dem Institute

Department of Political Science University of Gothenburg Sprängkullsgatan 19, Box 711 405 30 Gothenburg

Sweden

E-mail: contact@v-dem.net

V-Dem Working Papers are available in electronic format at www.v-dem.net.

Copyright ©2021 by authors. All rights reserved.

Walking the Talk: How to Identify Anti-Pluralist Parties ^∗

Anna Lührmann, Juraj Medzihorsky, and Staffan I.Lindberg V-Dem Institute

Department of Political Science University of Gothenburg

∗We are grateful to Kyle Marquard and participants of the HSE Faculty of the Social Sciences Seminar Series (12/2020), the University of Hamburg (12/2020) and Gothenburg (1/2021) for helpful feedback and comments. This research was supported by Vetenskapsrådet [grant number 2018-016114], PI: Anna Lührmann, Knut and Alice Wallen- berg Foundation to Wallenberg Academy Fellow Staffan I. Lindberg, Grant 2018.0144, and European Research Council, Grant 724191, PI: Staffan I. Lindberg, as well as by internal grants from the Vice-Chancellor’s office, the Dean of the College of Social Sciences, and the Department of Political Science at University of Gothenburg. We thank Ana Laura Ferrari, Ana Good God, Palina Kolvani, and Sandra Grahn for skillful research assistance. This paper partially builds on Lührmann et al. (2020).

Abstract

The recent increase of democratic declines around the world – “the third wave of autocratization”

– has sparked a new generation of studies on the topic. Scholars agree that these days the main

threat to democracy arises from democratically elected rulers, who gradually erode democratic

norms once in power. Is it possible to identify future autocratizers before they win power in elec-

tions? Linz (1978) and Levitsky and Ziblatt (2018) argue that a lacking commitment to democratic

norms reveals would-be autocrats before they reach office. Such anti-pluralist traits include de-

monizing rhetoric, the encouragement of political violence, disrespect for minority rights, and

lacking commitment to the democratic process. Comparative political science researchers have

not systematically collected and tested these potential early-warning indicators. This paper makes

use of a new expert-coded data set on virtually all relevant political parties worldwide from 1970

to 2019 (V-Party) to provide the first systematic empirical test of this argument.

1 Introduction

The vanguard of democracy’s contemporary enemies are paradoxically elected leaders and par- ties that once in power, erode democratic rights and institutions (Bermeo, 2016; Boese et al., 2020;

Lührmann and Lindberg, 2019; Svolik, 2015). Once autocratization gets underway, democracies stand only about a 20% chance to avoid breakdown and democracy has broken down in 36 coun- tries since the 1990s (Boese et al., 2020), mostly at the hands of parties and leaders once professing to be democrats. The key to democratic resilience that now preoccupies academic and policy com- munities is how to prevent autocratization to begin with. A first step is to answer the question:

Can we identify the traits of political parties that erode democracy once in power?

Using a unique dataset on 1,943 political parties across 1,759 elections in 169 countries from 1970 to 2019 (Lührmann, Düpont, et al., 2020), this paper shows that we can identify such par- ties with surprisingly high precision. The traits that signify would-be autocratizers are the four characteristics suggested in Linz’ canonical contribution from 1978: A rhetoric that is only “semi- loyal” to democratic norms, values, and institutions (Linz, 1978).

We provide a conceptualization of those traits as anti-pluralist, that signals of future authoritarianism. We show how this differs from other suggestions such as populism (Mudde and Kaltwasser, 2013; Müller, 2017; Norris and Inglehart, 2019) and ideology, and demonstrate with the first ever systematic test of how well the traits of these three partially competing ideas identify seemingly democratic but will-be-autocratic parties before they come into power.

This approach in this paper follows a long tradition in comparative politics emphasizing that political elite actors rather than structural conditions decide the fate of democracy. The work of Linz (1978) is canonical. Similarly, Bermeo (2003) argues cogently that elite action—not mass attitudes and behavior—were the final bow to every major case of democratic collapse in the 20

century. More recently, Mainwaring and Pérez-Liñán (2013) also stress that elite actors are key to explaining the ups and downs of democracy in Latin America in the post-Second World War era.

The approach in this paper follows a long tradition in comparative politics emphasizing that political elite actors rather than structural conditions decide the fate of democracy (e.g. Bermeo, 2003; Linz, 1978; Mainwaring and Pérez-Liñán, 2013). Political parties are the key actors in a democratic system (Aldrich, 2011). In his classical work, Richard S. Katz (1980, p.1) notes that the

“character of the parties in a political system is intimately related to the quality of its democracy”, and correspondingly, Ziblatt (2017) demonstrates that it was the strategic decisions of (conserva- tive) parties that sealed the fate of democracy in the Interwar years. Here we focus on the charac- teristics of political parties in the current period, recognizing that when a strong leader is present her/his identity will naturally shape the identity of the party.

The paper first discusses which traits of political parties are related to autocratization and develop the argument for anti-pluralism. Second, we introduce the V-Party data set and the rel-

1 Based on this argument, Levitsky and Ziblatt,2018later developed a list of early-warning indicators for potential anti-democratic behaviour.

evant measures of party characteristics. Third, we demonstrate a strong relationship between between levels of anti-pluralism in parties and autocratization if they come into power, before finally concluding.

2 Populism, ideology, or anti-pluralism

Some 35 years after the 1974 Carnation revolution in Portugal started the “third wave of democra- tization”, Putin rose to Prime Minister in president Yeltsin’s democratically elected administration on 16

August 1999. He was soon duly elected president (26

March 2000) with 53% of the vote in relatively free and fair elections. Although none of us knew it at the time, the “third wave of autocratization” (Lührmann and Lindberg, 2019) was getting underway. As a marker of what was going to come, the democratization process in Portugal’s famous revolution started off with a military coup, while the autocratization in Russia took off with a democratic election.

The ensuing descent into an electoral autocracy in Russia foretold a feature characterizing the present: Democratically elected parties and their leaders emasculate democratic norms and institutions to curtail competition to stay in power (e.g. Boese et al., 2020; Cassani and Tomini, 2018; Diamond, 2020; Haggard and Kaufman, 2016; Levitsky and Way, 2015; Norris and Inglehart, 2019; Plattner, 2015). This differs from most autocratization processes in the 20

century that originated in unconstitutional changes such as military coups, foreign interventions, and auto- golpes (Bermeo, 2016).

A recent study of the past 120 years shows that when a party seeking to derail democracy comes into power and autocratization gets on its way, there is only about a 20% chance to avoid breakdown (Boese et al., 2020). It is therefore critical to identify the traits of political parties that erode democracy once they come into power. That would enable both an “early warning system”

and open up for studying such parties before they take power to seek explanations why they gain ground, or not. It would also be instrumental for international democracy support community as a tool in democracy protection activities.

Signifiers of political parties that derail democracy

We find three sets of arguments in the literature on the traits signifying parties and their leaders

that derail democracy if they come into power. Some scholars argue that populism is the main

culprit (e.g. Mudde, 2004; Müller, 2017), while a second group suggests that qualities associated

with certain ideologies — for instance far-right or far-left — are the identifiers (e.g. Eatwell and

Mudde, 2003; Levitsky and Roberts, 2011). The third strand focuses on the commitment of political

leaders and parties to democratic norms and thus their degree of “illiberalism” or anti-pluralism

(e.g. Pappas, 2016; Zakaria, 1997). While there is some conceptual overlaps and the three are there-

fore not completely mutually exclusive, we argue that neither the core attributes of populism nor

of those of ideologies are the most important signifiers of parties and leaders that undermine

democracy once in power. Rather, we concur that lacking commitment to democratic norms as such - anti-pluralism is the best predictor of autocratization.

Populism

Pundits sometimes equate populism explicitly with authoritarian leanings (e.g. Galston, 2017; Müller, 2017; Plattner, 2010).

Müller (2017) suggests that the populist claim of representing ‘the people’ is always a threat to democracy. Others see populism as a threat to democracy because they connect it to corruption, the suppression of a critical civil society, and an exclusionary division of peoples into ‘us’ and ‘them’ (e.g. Galston, 2018). Yet, scholars have attributed populism to various political parties that claim to represent ordinary people’s vs the elites. They usually agree on three core ideas as characteristic of populism: people-centrism, anti-elitism, and an antagonism between the

“virtuous people” and the “corrupt elite” (e.g. Hawkins, 2010; Mudde, 2004; Rooduijn, 2014).

Based on such thin definitions, populism as such is not necessarily antidemocratic (see also Pappas, 2016) since not all leaders of populist parties seek to repress individual liberal rights and freedoms. In empirical analyses some researchers therefore distinguish between populist rhetoric and antidemocratic traits. For example, Norris and Inglehart (2019, p.4, 7) make a distinction between populism as an anti-elitist rhetoric, and authoritarianism as denying “liberal autonomy for the individual". Similarly, Rydgren (2017) opposes generalizing all new extreme-right parties as “populist”, and Akkerman (2003) finds that only populist parties that press for radical reform to benefit a popular majority are a threat, while populists who respect institutional pluralism are not. Populist parties can be progressive, conservative, socialist, authoritarian, or other in the two- dimensional economic-GALTAN space (Akkerman and Rooduijn, 2015; Norris and Inglehart, 2019, p.4).

We agree that it makes sense to distinguish populist rhetoric from commitment to democratic norms. Populism is defined by anti-elitist and people-centric rhetoric, predicated on a view of the “good” people and the “evil elite” (Hawkins, 2009). This is at the core of populist rhetoric and the common denominator of most scholarly definitions of populism (e.g. Akkerman, 2003;

Hawkins, 2009; Mudde, 2004; Müller, 2017; Norris and Inglehart, 2019; Rooduijn, 2014, p.543).

While populist rhetoric might correlate somewhat with autocratization, populism as such is not necessarily incongruent with commitment to democratic norms.

2 See also Cas Mudde and Jan-Werner Müller in Uri Friedman’s article in the Atlantic: https://www.

theatlantic.com/international/archive/2017/02/what-is-populist-trump/516525/ And Kurt Wey- land in the Washington Post: https://www.washingtonpost.com/news/monkey-cage/wp/2017/08/11/

there-are-4-big-barriers-to-the-the-populist-model-in-america-and-your-democracy-is-safe/

3 Mudde and Kaltwasser (2013, p. 150-151)—the protagonists of this ideational approach to populism, define it as follows:“Populism is a thin-centered ideology (...) with three core concepts : the pure people (depicted as a homogeneous and virtuous community), the corrupt elite (seen as a homogeneous but pathological entity) and the general will (idea that all individuals of a given community are able to unify their wills with the aim of proclaiming popular sovereignty as the only legitimate source of political power).” Similarly, Norris and Inglehart (2019, p.66) view populism as a

“claiming that (i) the only legitimate democratic authority flows directly from the people, and (ii) established power- holders are deeply corrupt and self-interested, betraying public trust”

Ideology

Political idelogy is an alternative suggestion to what threatens democracy (Akkerman, 2003; Mudde and Kaltwasser, 2013, p.38): Parties holding extreme opinions and advocating extreme measures, located farthest from the political centre that are juxtaposed to “normal” parties and thus sug- gesting severely radical policies. Two dimensions of party systems are used to identify extreme political ideologies. The classic left-right dimension and the more recent progressive-traditional dimension, or the GALTAN scale.

On the traditional left-right scale, focusing on economic policy, one argument is that the extreme-left is a threat to democracy by virtue of its fervor for state interventionism that facili- tates incumbent hegemony (Schamis, 2006; Weyland, 2013). One find such instances of chief exec- utives implementing redistributive policies while eroding democratic norms (e.g.Ecuador, Bolivia, and Venezuela) (Levitsky and Roberts, 2011, p.399). Historically, also extreme-right, conservative parties opposed democracy but we have not found empirical studies suggesting that current par- ties with far-right economic policies are likely to erode democracy.

There is also the GALTAN dimension, meaning that Green, Alternative, and Libertarian par- ties are located on the progressive side, while Traditional, Authoritarian, and Nationalist parties are on the other side of the spectrum (Bakker et al., 2015; Hooghe, Marks, and Wilson, 2002; Marks et al., 2006; Rohrschneider and Whitefield, 2009). Progressives favor expanding personal free- doms, believe the government should not interfere in people’s personal decisions, and oppose discrimination against minorities such as LGBTI, racial and ethnic groups, and others. Tradi- tionalists see the government as a moral authority that should guarantee order, favor traditional values above unbridled freedom, and value the national community often associated with a reli- gious affiliation (Hooghe, Marks, and Wilson, 2002). The threat to democracy is expected from from ideologies of the far-right traditionalists. For example, using a pooled cross-sectional de- sign and data from 30 European countries from 1990 to 2012, Huber and Schimpf (2017) show that traditionalist right-wing parties have negative effects on minority rights, and it is argued that such parties are associated with the propagation of violence and racism, which is detrimental for democratic stability (Koopmans, 1996).

Yet it seems to us that that political extremes may pave the way for, yet is not equal to an “at all

costs” approaches that becomes more dangerous to democracy the more convinced its supporters

get. With data on over 1,700 powerful political actors in 20 Latin American countries over a span

of 66 years, Mainwaring and Pérez-Liñán (2013) show that privileging policy goals over process

among powerful political actors is a key factor for democratic survival. They conclude that “com-

petitive regimes are highly vulnerable to breakdown if the most powerful actors are indifferent

to liberal democracy’s intrinsic value” (Mainwaring and Pérez-Liñán, 2013, p.135). Also ideologi-

cally relatively moderate parties can be a danger to democracy if and when the goal is considered

more important than the means. The key is that when ideology beats adhering to process and

norms, democracy is at risk. Recent experimental work shows that citizens who feel represented

by the executive are more willing to delegate the president more authority even at the expense of democratic principles, and the magnitude of the effect increases with partisan attachment (McCoy, Simonovits, and Littvay, 2020; Singer, 2018; Svolik, 2018).

Arguably, political ideology on either the classic or the cultural dimension may lead to demo- cratic decay if, and only if, political actors and voters place their ideological interests above demo- cratic norms. It is less ideology at the extremes as such than prioritizing goals over democratic norms that is the key factor. Therefore, we turn to the third approach.

Anti-pluralism

Building on the reasoning above and insights by Linz (1978), Levitsky and Ziblatt (2018) and Müller (2012), it becomes clear that future autocratizers are signified by lacking commitment to demo- cratic institutions, procedures, and norms. However until now, we lacked a robust conceptualiza- tion, data, and systematic empirical tests of which traits identify future autocratizers before they come to power.

The term illiberalism recently gained prominence in part due to its frequent use by Prime Min- ister Orbán in Hungary, who declared that he wants to transform his country into an “illiberal democracy”.

Among others, scholars like Zakaria, 1997 and Pappas (2016, p. 31) use “illiberalism”

to denote a system of government that holds multiparty elections but does not protect basic liber- ties, and “illiberals” for the parties and leaders seeking to derail democracy. But these terms suffer from frequent misinterpretations and thus the risk of a collective conceptual confusion. “Liberal”

with its Latin root lïberälis has many meanings, from negative connotations of unrestrained by prudence, to open-mindedness; broad-based education favoring independent thinking; state non- interventionism in the economic sphere and beyond; favoring of social or even socialist reforms;

and designated political parties with varying agendas.

Instead, consider that most relevant definitions of democracy rest on the foundation of plural- ism in the full sense of the word (Dahl, 1971), which goes far beyond simplistic understandings of a mere plurality of parties or interest groups. As discussed for example by Sartori, 1997, pluralism is a value system itself positing not only tolerance but also respect for opposing views on the basis of mutual reciprocity. The limits to tolerance is only harm to the principle of tolerance itself, as in Popper’s words: “tolerate[s] all who are prepared to reciprocate, i.e.who are tolerant” (Popper, 1945, p. 293). Pluralism thus requires consensus about the principles of reciprocal tolerance, which in effect makes democracy not “the peaceful resolution of conflict” but rather makes disagreements something less than conflict. Pluralism and its principle of tolerance also informs the democratic principle that while the majority may rule, its legitimate course of actions are limited—they must respect the rights of minorities. Thus, pluralism embraces “that societies are composed of several social groups with different ideas and interests.” (Mudde and Kaltwasser, 2013, p. 152).

Several authors (Mudde and Kaltwasser, 2013; Müller, 2017; Norris and Inglehart, 2019, p.51)

4 Orbán in July 2014 as quoted by Plattner (2019, p. 9).

view pluralism as the opposite pole of populism, yet anti-pluralism does not feature in their defi- nitions of populism. That seems to be like removing the active ingredient. In scholars’ and in the everyday sense of “populism", it does not necessarily include anti-pluralism. For analytical clarity, it is therefore critical to separate the two terms.

Anti-pluralism has four key characteristics. The first one is unwillingness to commit to the democratic process as legal means for gaining power. Dahl (1971) formulated the minimum require- ments in terms of institutional guarantees safeguarding true pluralism in competition among po- litical parties for power, and enabling orderly alternations. When parties do not commit to respect for these institutions regulating the means to access power, it signals anti-pluralism.

The second attribute is denial of the legitimacy of dissenting parties and opponents. This follows directly from the principle of reciprocal tolerance on which pluralism rests. This is well exem- plified by John McCain’s remarks during the 2008 presidential campaign about his political rival Barack Obama: “He’s a decent family man, a citizen that I just happen to have disagreements with on fundamental issues, and that’s what this campaign is all about.”

If political actors instead dele- gitimize, severely personally attack, or demonize their opponents, it is an indication of lacking commitment to pluralism (Levitsky and Ziblatt, 2018, p.23-24). This includes dehumanizing oppo- nents or describing them as an existential threat, subversive, criminal, or foreign agents. The only exception is that one may legitimately deny the rightful existence of parties that do not sign on to the consensus of the principle of pluralims itself. Popper (1945, p.130) pointed to this in his work on the paradox of freedom. If freedom includes the right to abolish itself, citizens are prohibited from exercising democratic freeddoms in the future. Parties are endorsing pluralism only as long as they are committed to the same democratic processes, institutions, and freedoms that made pluralism possible in the first place. Consequently, anti-pluralist democracy is an oxymoron.

The third key feature is toleration or endorsement of the use of political violence. Pluralism is pred- icated on that disagreements with political opponents should be solved according to the agreed rules and procedures (Sánchez-Cuenca, 2003, p. 69). One of the key components of the rule of law is the principle that the law protects civil liberties from arbitrary violation by the state even if elected representatives would agree to such infringements (Merkel, 2004, p. 39). Thus, a key indication for a weak commitment to the rule of law is support for the notion that the will of the majority should be implemented even if doing so would violate the physical integrity of op- ponents. In the extreme, this entails the threat or use of political violence. Thus, toleration or endorsement of the use of political violence is a clear signal of an anti-pluralist party or leader (Levitsky and Ziblatt, 2018, p. 23-24).

Finally, indications that a party and its leaders could consider curtailing the civil liberties of minor- ity groups is a clear signal of anti-pluralists. Dahl (1971) and others rightly emphasize that democ- racy requires not only choice in elections, also a plurality of information and opinions that can be expressed freely. Civil liberties enable pluralism by providing the individual with the rights to life, liberty, and economic activity but also by assuring citizens the opportunity to express themselves,

5 https://time.com/4866404/john-mccain-barack-obama-arab-cancer/

associate with others, assemble, protest, and be informed by varying viewpoints. Civil liberties are fundamental prerequisites for pluralism. To illustrate this point, consider the following “chain of democratic choice”.

If citizens are unable to express themselves others cannot learn about their preferences. Even if their preferences are known, if citizens are not allowed to form organiza- tions their preferences are not aggregated and amplified at the societal level. Even if a variety of associations is permitted, if the media is either not willing or not permitted to report on them citizens are not presented with the existing plurality of options. Thus, without civil liberties citi- zens do not have a real choice on election day, and we do not know if their vote or opinion polls reflects actual preferences. A system in which the people’s preferences are not known, cannot be described as being ‘ruled by the people’ and hence is not democratic. Thus, if pluralism is severely undermined in any piece of the democratic chain it breaks, and no meaningful democracy exist.

We submit that these are four key issues where anti-pluralist parties may reveal themselves to be likely autocratizers if and when they come into government. Anti-pluralist parties lack com- mitment to i) the democratic process as the legal means of gaining and losing power; ii) the legitimacy of political opponents; iii) peaceful resolution of disagreements and rejection of political violence; and iv) unequivocal support for civil liberties of minorities. We define those political parties as to a varying degree anti-pluralist that register lack of commitment in these areas.

The litmus test identifying anti-pluralists

The reasoning above coheres with Linz’s (1978, p.& 29) famous “litmus test” of what characterizes political actors disloyal to the democratic system (see Table 1). Levitsky and Ziblatt (2018) build on Linz’s work and argue that one can identify challengers to democracy by examining their public statements before they come to office. They provide a list of four indicators that are very similar to Linz’s original catalogue: 1) A weak commitment to democratic rules of the game; 2) The denial of the legitimacy of political opponents; (3) The toleration or encouragement of violence; and (4) A readiness to curtail civil liberties. They also detail further specifics of instances that fall under each category. Müller (2012) suggests in a similar fashion that dispension by parties and their leaders of each of the following four political positions, constitutes an attack on “core democratic principles” and in effect are anti-pluralist: (1) Exclusion of groups of citizens; (2) denial of the

“dignity” of groups of citizens; (3) positive identification with past “ethnic cleansing or genocide”;

and (4) denial of pluralist society by claiming to “speak in the name of the people as a whole”

. His four items map well onto the previous ones but are more narrowly focused.

What remains is how these four indicators could be measured. When considering suitable indi- cators we must be cognizant that anti-pluralist parties typically seek to shore up their credentials as “regular democratic parties” in official party documents. For example, data from the Mani- festo Project (Volkens et al., 2019) show that such parties routinely pay lip service to democracy as the “only game in town”. The German far-right party AfD scores high (4.08) and the center-right

6 See Schedler (2002, pp. 39-41) for the development of similar notion focusing on election quality.

7 Müller (2012, p. 1287).

Table 1: Indicators of anti-pluralist political actors

Linz (1978) Levitsky and Ziblatt (2018) Lührmann et al. (2020)

Unwillingness to publicly commit to legal means for gaining power

Expresses willingness/need to violate the constitution;

Expresses sympathy for non-constitutional means of accessing power;

Attempts to undermine the legitimacy of elections

Low commitment to the democratic process

Denial of the legitimacy of democratic political parties to participate in political processes

Describes rivals as subversive, criminal, or foreign agents;

Claims that rivals constitute existential threat

Demonization of political opponents

No rejection of the use of force;

Willingness to ask for the armed forces

Has ties to armed gangs or militias;

Sponsors or encourages mob attacks on opponents; Endorses or praises political violence

Encouragement of political violence

Curtailment of the civil liberties of democratic parties’ leaders and supporters

Supports laws or policies restricting civil liberties;

Threatens to take legal action against critics; Praises repression

Disrespect for fundamental minority rights

Source: Levitsky and Ziblatt (2018), Linz (1978), and Lührmann, Düpont, et al. (2020)

party CDU low (0.74) on their measure counting the favorable mentions of democracy in the party manifestos. Using public party documents is therefore not likely to create a valid measure.

But rhetoric in campaign speeches and other events directed at party supporters and others can be more authentic and revealing (Maerz and Schneider, 2021). Maerz and Schneider (2019) compared 4,740 speeches from heads of government in 27 countries between 1999 and 2019, and found that leaders in autocratic countries use a substantially less democratic style of rhetoric than leaders of democratic countries. Linz (1978) found that political parties later seeking to break down democracy, are typically explicit with their true anti-pluralism while they are still not in influential positions. Once their appeal expands to a broader audience, they are likely to “convey equivocal messages in order to maintain their radical opposition to the system while claiming to aim at a legal access to power” (Linz, 1978, p. 29). Thus, we should expect such parties to come with increasingly vague anti-pluralist assertions until they accede to power and then embark on a path of autocratization (Levitsky and Ziblatt, 2018). There are exceptions such as Fidesz’s rhetoric during their 2010 return to government shows.

Even so, it is important to have sensitive

8 https://www.economist.com/eastern-approaches/2010/07/04/read-the-large-print

measures capturing even if some actors openly display only a mild anti-pluralist tendency.

But relaying on speech data from party leaders’ public appearances would be limiting since not all campaigns were covered equally extensively. Coverage is going to be biased towards more developed countries with large media infrastructures, as well as towards later years when more and more materials are available for these techniques. So while studies like Maerz and Schneider (2019) are informative in terms of contemporary analysis and for hypothesis-generation, the data is less useful for the purposes here.

We therefore turn to the new, unique V-Party data set (Lührmann, Düpont, et al., 2020) build- ing on V-Dem’s gold-standard methodology for multiple expert-based ratings (Coppedge et al., 2020; Pemstein et al., 2020). A selection of V-Party’s indicators were designed to capture Linz’s lithmus-test and therefore fit for the purposes here:

• Low commitment to the democratic process (v2paplur): “Prior to this election, to what extent was the leadership of this political party clearly committed to free and fair elections with multiple parties, freedom of speech, media, assembly and association?”

• Demonization of political opponents (v2paopresp): “Prior to this election, have lead- ers of this party used severe personal attacks or tactics of demonization against their oppo- nents?”

• Disrespect for fundamental minority rights (v2paminor): “According to the leader- ship of this party, how often should the will of the majority be implemented even if doing so would violate the rights of minorities?”

• Encouragement of political violence (v2paviol): “To what extent does the leadership of this party explicitly discourage the use of violence against domestic political opponents?”

3 Data and descriptive empirical analysis

The V-Party dataset was constructed by 665 purposefully selected country experts assessing the identity of all political parties with a vote share of more than 5% in a legislative election between 1970 and 2019 in their country of expertise, across 169 countries. The dataset covers 1,943 politi- cal parties across 1,759 elections—or in total 6,321 party-election year units. Typically at least 4 experts contributed to each question.

All items are measured on a five-point ordinal scale and aggregated to latent, interval measures while correcting for possible between-expert differences in scale use, using V-Dem’s custom-designed IRT model (Pemstein et al., 2020). We have rescaled

9 Coder were shown the following clarification: “This concerns the rights enshrined in the Universal Declaration of Human Rights, which apply to everyone ‘without distinction of any kind, such as race, colour, sex, language, religion, political or other opinion, national or social origin, property, birth or other status.’ The declaration protects—among others—freedom of speech, property, religion, peaceful assembly and association.”

10 Coders were shown the following clarification: “ ‘Domestic political opponents’ refers to all political opponents, with the exception of those who are engaged in an armed conflict with the state. They may be other political parties or other political groups and movements.” (Lührmann, Düpont, et al.,2020)

11 Mean = 4.55; median (p50) = 4; p25= 3; p10= 2.

the variables here so that high values indicate a high level of anti-pluralism. For details see ap- pendix A.

Index aggregation

We aggregate the four discussed indicators to the Anti-Pluralism Index (API). They do not cap- ture equally severe violations of democratic processes, institutions, and norms. In particular, an explicit denial of democratic institutions and processes such as elections is a more severe rejection of pluralism than harsh language towards opponents. The first one captures an openly authoritar- ian trait whereas the latter one might also express a contentious political culture. Based on this reasoning, we compute the index as a transformed weighted average of the input indicators using the following formula:

v2xpa_rival

= 1 − Φ

( 0 .5 × v2paopresp

+ 2 × v2paplur

+ v2paminor

+ v2paviol

4 .5

) ,

where 𝑖 indexes observations, Φ is the standard normal cumulative density function, and the four indicators are Demonizing opponents (v2paopresp), Low commitment to democratic processes (v2paplur), Disrespect for fundamental minority rights (v2paminor), and Encouragement of polit- ical violence (v2paviol). Figure B.1 in the Appendix shows the joint distributions of the API and its components.

Data validation and descriptive analysis

It is usual to asses all three forms convergence, content, and construct validity for a new mea- sure (Adcock and Collier, 2001). Since no other measure has yet captured the anti-pluralist traits of political parties, we assess convergent validity by comparing the values on the Anti-Pluralism Index of ruling parties in democracies (top of Figure 1) with the one in autocracies (bottom). Re- assuringly, this shows a stark difference between ruling parties in these two regime categories. In democracies, the smoothed median score in 2019 was 0.28 and in autocracies much higher with 0.85. The time trend is also informative. The median governing party in democracies has become more anti-pluralist in recent decades, with the smoothed median score rising from 0.08 in 1970 to 0.28 in 2019, which is congruent with research showing that contemporary threats to democ- racy typically come from within the government (Bermeo, 2016). The median governing party in autocracies has become somewhat less anti-pluralist in the same period, reflecting the mimicking of multi-party elections in most autocracies (Schedler, 2002). This finding thus supports the not only the convergence but also the content validity of the new index.

To further assess the content validity of the API, we explore some relevant cases. Figure 2 shows the movement of the US Republican and Democratic parties in this century on two dimensions:

anti-pluralism and left-right positioning on economic policy, with other parties positioning at

the last election in relief. The Republican party has not changed left-right placement but moved

Figure 1: Anti-Pluralism Index (API) of governing parties. Parties that gained or retained the Head of Govern- ment post in national elections. The lines with 95% regions estimated with a quantile GAM with Gaussian Process over days. Elections split by regime (v2x_regime): autocracies{0, 1}, democracies {2, 3}. Color by economic left-right on the OSP scale: left[0, 2], center (2, 4], right (4, 6].

strongly in an anti-pluralist direction with a score on the API of 0.69. This reflects that Trump in his 2016 presidential campaign made personal, demonizing attacks on political opponents leading to a high score of 0.86 on the indicator on demonization of the opponent (see Table 2).

He also condoned violence towards his political opponents, saying about Clinton that “If she gets to pick her judges, nothing you can do, folks [...] Although—the second amendment people—maybe there is, I don’t know”,

and towards protesters at his rallies, saying things like “I loved the old days, you know what they used to do to guys like that when they were in a place like this? They’d be carried out on a stretcher, folks and even offering to relieve the consequences of political violence: “ [...]knock the crap out of them, would you? [...] I will pay for the legal fees, I promise”.

This is reflected in a high

12 https://www.theguardian.com/us-news/2016/may/25/donald-trump-crooked-hillary-clinton- nickname-ignore-fight

https://www.theguardian.com/us-news/video/2016/aug/12/donald-trump-hillary-clinton-day- wrong-florida-video

https://abcnews.go.com/Politics/crooked-hillary-marco-donald-trumps-nicknames/story?id=

39035114

13 https://www.theguardian.com/us-news/2016/aug/09/trump-gun-owners-clinton-judges-second- amendment

https://www.bbc.com/news/election-us-2016-37036856 The second amendment to the Bill of Rights in the constitution of the United States of America has long been subject to controversial debate and guarantees that “...the right of the people to keep and bear Arms, shall not be infringed”

https://www.constituteproject.org/constitution/United_States_of_America_1992?lang=en.

14 https://www.theguardian.com/us-news/video/2016/feb/23/donald-trump-says-hed-like-to-punch- protester-at-nevada-rally-video

score for the encouragement of violence (0.35, Table 2). From the Democratic party no similar statements are known.

Figure 2: The movements of the two major US parties on the Anti-Pluralism Index and economic left-right since 2000.

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

Economic Left-Right Anti-Pluralism

2000 2018

2000 20022004

20062008 2010

2012 2014 2016 2018

Democrats

Republicans

SPD

Germany 2017

CDU

Germany 2017

PSOE

Spain 2019

PP

Spain 2019

Labour

UK 2019

Conservatives

UK 2019

+ Autocracies

Median 21^st century ruling party

+ Democracies

Median 21^st century ruling party

AKP

Turkey 2015

BJP

India 2019

Fidesz

Hungary 2018

PiS

Poland 2019

SYRIZA

Greece 2019

MAS

Bolivia 2019

PAIS

Ecuador 2017

SPUV

Venezuela 2015

PSL

Brazil 2018

By 2016, the rhetoric of the Republican party was in the sense more similar to autocratic rul- ing parties such as the Turkish AKP (1.0) and Hungarian Fidesz (0.88) than to typical center-right governing parties in democracies such as the Conservatives in the UK (0.35) or CDU in Germany (0.05). This also applies to Bolsonaro (PSL, Brazil) whose rhetoric during the 2018 presidential campaign was filled with demonization of the opposition and violence promotion (Hunter and Power, 2019). During a campaign event, while pretending to hold a riffle, he told supporters: “Let’s shoot the Workers’ Party supporters”, which his campaign later dismissed as a joke.

After a wave of violence against minorities carried by his supporters, Bolsonaro commented “What has that

https://www.bbc.com/news/election-us-2016-35793103

15 https://www.nytimes.com/2018/10/17/world/americas/brazil-attacks-bolsonaro-haddad.html

got to do with me?”.

These positions are reflected on PSL’s score of 0.95 on the API. Likewise, PiS, the ruling-party in Poland since 2015, had a high API of 0.80 at the 2019 elections reflecting for instance that the government-controlled media attacked and demonized PiS’ opponents and accused them of threatening traditional Polish values (Markowski, 2020). A more drastic example is the Turkish AKP, which receieved the highest possible score of 1.0 ahead of the 2018 elections reflecting that not only the party used anti-pluralist rhetoric, but also that the AKP-controlled government interfered with the campaign, jailing political opponents and controlling access to national television.

The Greek SYRIZA is on an intermediary level in terms of anti-pluralism with an overall score of 0.53 ahead of the 2015 election. Their overall discourse was inclusionary, upholding human rights, which resulted in low scores on commitment to the democratic process (0.18). However, similarly to other populist parties in Europe, SYRIZA criticized opponents of being subservient to foreign powers and the international banking system (Aslanidis and Rovira Kaltwasser, 2016;

Font, Graziano, and Tsakatika, 2021; Lisi, Llamazares, and Tsakatika, 2019) and, thus, did not perform well on demonization of political opponents (0.60). Similarly, the Conservative Party in the UK has moved somewhat towards anti-pluralism due to severe instances of demonization of the opposition during the 2019 electoral campaign reflected in a high API score of 0.35. However, overall they are still in the pluralistic spectrum.

From the leadership of Angela Merkel’s CDU no anti-pluralistic statements or behavior is known, which corresponds to a low score on the API of 0.05 in 2017.

Table 2: Scores of selected parties on the Anti-Pluralism Index and its component indicators. Indicators have reversed scales to range from 0 (pluralist) to 1 (anti-pluralist).

API Component Indicators

Party Country Election Anti-

Pluralism Index

Demo- cratic Process

Demo- nization

Disrespect minori- ties

Encourage- ment of Violence

AKP

Turkey 2018 1 0.96 0.99 0.93 0.95

PSL Brazil 2018 0.95 0.53 0.91 0.88 0.53

Fidesz Hungary 2018 0.88 0.41 0.96 0.79 0.38

PiS Poland 2019 0.80 0.35 0.91 0.71 0.37

Republicans USA 2016 0.69 0.28 0.86 0.61 0.35

SYRIZA Greece 2015 0.53 0.18 0.60 0.49 0.52

Conservatives UK 2019 0.35 0.09 0.87 0.82 0.04

Democrats USA 2016 0.13 0.13 0.37 0.31 0.02

CDU Germany 2017 0.05 0.05 0.15 0.27 0.04

a In 2018 the AKP ran as part of the People’s Alliance.

16 https://www.theguardian.com/world/2018/oct/11/brazil-election-violence-bolsonaro-haddad

17 https://www.nytimes.com/2018/06/22/world/europe/turkey-elections-erdogan-opposition.html

18 https://www.independent.co.uk/news/uk/politics/conservative-social-media-strategy-attack- jeremy-corbyn-labouir-twitter-facebook-video-stormzy-akala-jme-a7784406.html

4 Anti-pluralism in action: autocratization after winning power

The idea here is that identifying anti-pluralism among parties before the come into power is indi- cation that they will pursue a path of autocratization if and when they rule. This notion rests on the assumption that future autocratizers tend to reveal their lacking commitment to democratic norms before taking power in speech acts. That parties that are inherently anti-pluralist would be expected to undermine democratic pluralism if and when they assume power, should be uncon- troversial. The question is if the Anti-Pluralist Index can capture that in advance, and the extent to which anti-pluralism in parties is a driver of autocratization.

In order to test these ideas about the relationship between the characteristics of political par- ties rhetoric before, and the behavior of political leaders after assuming office, we conduct de- scriptive analysis and run regressions on a subgroup of 771 cases: Political parties, which hold the office of Head of Government after a given election in a democracy.

As we detail above, these ideas reflected longstanding hypotheses, which have not been susceptible to empirical testing due to lack of data. Thus, our analysis here are both relevant for their own sake as well as an exer- cise of construct validation (Adcock and Collier, 2001). But first we introduce how we measure autocratization.

Operationalizing autocratization

Our dependent variable is binary, taking the value of 1 if the country is undergoing substantial autocratization in the year after the elections and 0 if it is not. This is based on autocratization episodes identified by Edgell et al. (2020). Such autocratization represents a decline in V-Dem’s Electoral Democracy Index (EDI) of more than 10% of the index value during one year or over a connected time period. Operationalizing autocratization in this way is superior to a simple year- to-year change as it allows us to capture substantial and gradual processes of autocratization, while at the same time not registering year-to-year fluctuations that may be do measurement noise.

We focus on autocratization led by elected incumbents as the most common mode of contem- porary autocratization—in contrast to autocratization starting with an illegal power grab such as military coups or international invasion. In prior research, instances of such incumbent-led autocratization have been termed “autogolpes” (Przeworski et al., 2000, p. 21), but here we also include the more gradual forms. We identified whether each autocratization episode was led by an elected leader or not using a range of qualitative data sources as well as Powell and Thyne (2011) and Coppedge et al. (2020).

19 To identify which party is in power we use an updated version of the variable v2pagovsup from Lührmann, Düpont, et al. (2020). To identify democracies we use the variable v2x_regime in the year before elections from Coppedge et al. (2020).

Descriptive analysis of the relationship between anti-pluralist ruling par- ties and autocratization

When anti-pluralist parties are in power in democracies 29% autocratize in the year after they won (or defended) the office of the Head of Government (see Group IV on Table 3).

Conversely, after pluralists win office, only 6% of democracies autocratize (Group II) while most remain as demo- cratic as they were before (94%; Group I). This is a much less likely outcome after the election of anti-pluralists (71%; Group III). The large Group I (democratic stability with pluralists governing) is not much of a surprise as one would expect pluralists to rule in a democracy. However, all other groups require further scrutiny.

Table 3: Anti-pluralists winning elections in democracies and autocratization (t+1), 1970-2018.

No autocratization Autocratization Pluralists

in power

542 (94%) 36 (6%) 578 (100%)

I: Democratic stability with plu- ralists governing

II: Autocratization with plural- ists governing

Anti- pluralists in power

137 (71%) 46 (29%) 193 (100%)

III: Democratic stability with anti-pluralists governing

IV: Autocratization with anti- pluralists governing

All 679 (88%) 92 (12%) 771

The center line in Figure 3 illustrates the relationship between anti-pluralism in ruling parties (x-axis) and the level of electoral democracy in their countries (y-axis). It shows that, typically, the more anti-pluralist ruling parties become, the lower the level of democracy. Political parties to the right of the vertical line (at the API value of 0.429) are considered as anti-pluralists here and the lines in color illustrate the trajectories of selected Group IV parties. The Polish Law and Justice Party (PiS), the Hungarian Fidesz Party and the Turkish Justice and Development Party (AKP) have started in this millennium with a score in the pluralist party spectrum. The Indian Bharatiya Janata Party (BJP) exhibited already in 1999 some level of anti-pluralism. However, over the course of the last 20 years, all four parties have become more and more anti-pluralists. Under their leadership, their countries have also becoming increasingly autocratic, with Hungary losing its status as democracy in 2018 and Turkey in 2014.

The Justice and Development Party (AKP) increased its anti-pluralist traits between every elec- tion since its foundation. When Erdoğan came to power after the 2002 election, he had promised reforms that would enhance the separation of powers, the independence of the judiciary, in- crease the freedom of press, and strengthen the rule of law. However, Erdoğan’s increasingly anti-pluralist actions continued to break his promises. Over the years, he has continuously show- cased highly anti-pluralist behavior such as cracking down on protesters and accusing them of

20 Parties scoring more than 0.429 on the API are considered here as anti-pluralist. This is the above the 75th percentile for governing parties in democracies in our sample.

Figure 3: Anti-pluralism of governing parties in election years and V-Dem’s Electoral Democracy Index in the first post-election year (𝑡 + 1).

Yes NoAutocratization episode

0% 25% 50% 75% 100%

0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0

Anti−Pluralism Electoral

Democracy Index (t+1)

2.5%

25%

50%

75%

97.5%

Smoothed conditional quantiles 2000●

MVR/PSUV

Venezuela 2015

●

●

2002 ●

AKP

Turkey 2015

● ●

●

●

1999

BJP

India 2019

● ● ●

●

2002

Fidesz

Hungary 2018

●

● ●● ●

2001

PiS

Poland 2019

● ●

Yes No Main governing party

Democracy Autocracy

Autocratization episode

Yes No

attempting to destabilize the country (Taspinar, 2014), orchestrating prosecutions of political op- ponents, and describing the separation of powers as “an obstacle” that he would overcome by hollowing out the judiciary (Karaveli, 2016).

The Polish Polish Law and Order’s (PiS) became gradually more anti-pluralistic throughout the years, going from an API score of 0.23 in 2005, when it was first elected governing party, to 0.71 when they returned to that position in 2015. This increase resulted from of a change to emphasizing nationalism, disrespecting fundamental minority rights, and demonizing opponents (Harper, 2010, p. 24). Additionally, PiS has undermined the independence of the judiciary, checks and balances, as well as freedom of expression (Markowski, 2020). This has led Poland’s EDI score to decline from 0.87 in 2014 to 0.72 in 2019.

In a similar turn of events, Hungary scored well on the EDI prior to the 2010 election, when Fidesz won the elections. The Orbán government has since placed constrains on civil society and restricted freedom of expression as well as academic freedom, passing legislation that led to the closing of the Central European University in Budapest.

Additionally, Orbán blatantly demo- nizes the opposition.

Thus, there has been a significant decline in EDI, with scores declining from 0.54 in 2014 to 0.32 by the time new elections were held in 2018.

The MVR/PSUV in Venezuela have consistently scored high on anti-pluralism with a score of 0.95 already in 1998, when Chávez won his first election. Venezuela then embarked on a dras- tic autocratization episodes with EDI scores dropping from 0.78 in 1999 to 0.35 by 2015. After Mauduro succeeded Chávez, PSUV was defeated in the 2015 legislative elections. In response,

21 https://www.ceu.edu/istandwithceu/timeline-events

22 https://www.nytimes.com/2018/04/07/world/europe/hungary-viktor-orban-election.html

Maduro stripped the parliament of power and designated the Supreme Court to take over the functions of the National Assembly. By consolidating the power in the executive branch, control- ling public institutions, and censoring and persecuting the opposition, PSUV has managed to stay in power for over two decades. (Alarcón, Álvarez, and Hidalgo, 2016; Corrales, 2020)

The Hindu-nationalist BJP became India’s governing party in 2014 led by Narendra Modi.

Democracy has deteriorated since then, with EDI scores declining from 0.71 to 0.54 between the 2014 and 2019 parliamentary elections. A series of policies have diminished freedom of expression and academic freedom, and repression of civil society has increased. BJP’s vision of India as a Hindu nation has also led to persistent discrimination against Muslims (Ganguly, 2020; Maerz, Lührmann, et al., 2020; Varshney, 2019). This is reflected in BJP’s 0.84 score on the API in 2019.

However, not all countries autocratize with anti-pluralists in office (Group III). In 137 cases, autocratization has not started in the year after an anti-pluralist was in power (lower rug on the right of Figure 3). The median API in that group is with 0.68 lower than in the group where anti-pluralism is followed by autocratization (0.77, Group IV). This may indicate that a certain threshold is required for autocratization to be the consequences, which is an issue for future research. A similar argument might apply for the level of liberal constraints on the executive which are with a median score of 0.73 higher in Group III than in the autocratizing group IV (0.66).

Parliamentary and judicial oversight as well as the rule of law may help to prevent autocratization.

24

A moderately anti-pluralist party that did not substantially erode democracy when ruling is the Italian Forza Italia with an anti-pluralism score of 0.72 in 2001. Silvio Berlusconi ruled Italy from 2001 to 2006 and from 2008 to 2011, and while he was known for anti-pluralist verbal provocations, the Italian institutions remained more or less intact (Verbeek and Zaslove, 2016).

A few elections (36) have been followed by autocratization even though a pluralist party was in office (Group II). In some a party used pluralist rhetoric while eroding democracy. For example, Citizens for European Development of Bulgaria (GERB) scored 0.13 on the API when gaining power in 2009. Yet, the quality of democratic institutions declined in Bulgaria from 2009 to 2019, as indicated by shift on the EDI from 0.72 to 0.59 (Maerz, Lührmann, et al., 2020, pp. 10-13). GERB has persued an agenda of state capture by putting loyal individuals in charge over media outlets, the Supreme Justice Council, and anti-corruption agencies providing de-facto impunity to top party members (Ganev, 2018). A similar but less pronounced autocratization process occurred in Chile under pluralist leadership from 2011 to 2019,

and in Israel (2010 to 2019).

In other cases, the pluralist party leader wining the elections was removed from office under dubious circumstances as with President Dilma Rousseff in Brazil in 2016 (Chalhoub et al., 2017). Finally, there are cases in this group registering only minor democratic declines under pluralist rule but more severe declines happen later under anti-pluralist rule, for instance in Hungary in 2007 and the United

23 Based on the Liberal Component Index at𝑡 − 1 from Coppedge et al. (2020).

24 For a similar argument see Boese et al. (2020).

25 https://rsf.org/en/news/rsf-alarmed-escalating-violence-against-reporters-chile

26 https://foreignpolicy.com/2010/03/10/civil-society-and-human-rights-in-israel-and- elsewhere/

States in 2015.

The discussion here illustrates the construct and convergent validity of out measures. First, the developments on the anti-pluralism index correspond to real-life events. Second, it captures anti-pluralist traits of ruling parties before they begin eroding democracy. Third, the data is nu- anced enough to show development over time; i.e.to not judge a party in 2000 by what it did later in the century. Fourth, they also show that the measures of autocratization and anti-pluralist rhetoric capture distinct concepts.

Regression analysis: research design

We investigate the relationships between autocratization episodes and senior government party characteristics with regression models. Our data contains 771 parties that have gained or defended the Head of Government post in an election in 104 democracies from 1970 until 2018. To capture possible nonlinearities, we use Generalized Additive Models (GAM) under which party charac- teristics as well as additional covariates are included via Gaussian Process (GP) smooths (see e.g.

Hastie and Tibshirani, 1990). Under a binary 𝑦-variable, the Gaussian GAM is a more flexible counterpart of the popular Linear Probability Model (see e.g. Angrist and Pischke, 2008), and the smooths may be interpreted analogically to slopes under the LPM.

Additional party characteristics

As discussed above, there are also other party characteristics that influence both the propensity of parties to take anti-pluralist positions as well as their inclination to commit autocratization.

We therefore include measures from the V-Party data set (Lührmann, Düpont, et al., 2020) on three party characteristics. First, we adjust for a thin version of populist rhetoric, which captures the extent to which parties use anti-elitist rhetoric and “glorify the ordinary people and identify themselves as part of them” (Lührmann, Düpont, et al., 2020, p.26). This index is aggregated using the harmonic mean, in order to allow for some substitutability between the input indexes, but only to a limited degree.

Second, research shows that political ideology associates with undemocratic behaviour, following an U-shape under which the incidence of undemocratic behavior is higher on both extremes than in the middle.

Therefore, we adjust for the traditional, or economic left-right dimension of the party system.

Third, we capture the distinction between progressive and conservative parties with an index on cultural characteristics, which aggregates measures on immigration (v2paimmig), LGBT Social equality (v2palgbt), Cultural superiority (v2paculsup),

27 A more detailed discussion is available in the Appendix.

28 See discussion above.

29 Experts were asked to “locate the party in terms of its overall ideological stance on economic issues” on a 0 (far-left) to 6 (far-right) scale. They were shown the following clarification: “Parties on the economic left want government to play an active role in the economy. This includes higher taxes, more regulation and government spending and a more generous welfare state. Parties on the economic right emphasize a reduced economic role for government: privatization, lower taxes, less regulation, less government spending, and a leaner welfare state” (Lührmann, Düpont, et al.,2020).

religious principles (v2parelig), and working women (v2pawomlab). For more details on these measures see Appendix A. When used in regressions, all party characteristics are scaled from 0 (left/non-populist) to 1 (right/populist).

Contextual covariates

To adjust for the socio-economic and political context in the pre-election year ( 𝑡 − 1) we include the following covariates. First, since the likelihood of autocratization as well as the election of anti-pluralist parties might vary with the level of liberal constraints in a democracy, we include V-Dem’s Liberal Component Index (LCI) (Coppedge et al., 2020). It captures both legislative and judicial constraints on the executive, and the rule of law. Democracies are said to be more likely to break down if they have a presidential system (Linz, 1978; Svolik, 2008). Thus, we adjust for presidential systems using V-Dem data.

Second, both the rise of parties which challenge democracy as well as autocratization is ob- served more frequently in weak economies (Przeworski et al., 2000) and during economic crises (Bernhard, Nordstrom, and Reenock, 2001). Democratic regime’s failure to combat economic crises in a way satisfactory to the citizens reduces the legitimacy of the incumbent regime and democracy, and thus increases the risk of autocratization following an economic crisis. This is especially true in countries with a background of poor economic performance, a lower level of development, or a relatively short democratic history (Haggard and Kaufman, 2016; Møller, 2018; Waldner and Lust, 2018 p. 323-324). Therefore, we adjust for GDP/capita (natural logarithm) and GDP growth/capita using data from the Maddison project

. An equal distribution of re- sources reduces the likelihood of autocratization (e.g. Diamond, 2015; Haggard and Kaufman, 2016;

Leininger, Lührmann, and Sigman, 2019; Tomini and Wagemann, 2018). Therefore, we adjust for inequality with V-Dem’s Equal Distribution of Resources Index, which measures the extent to which resources are equally distributed in a society.

Finally, it seems plausible that autocratization becomes more likely in a global climate of many reverse trends and less likely in a context of democratization (Lührmann and Lindberg, 2019).

Therefore, we adjust for the share of countries going through autocratization and democratization episodes each year.

For similar reasons, we adjust for the average regional EDI, always excluding the country of observation from the average (Coppedge et al., 2020). We also include year to account for temporal effects and allow for non-linear relationship with a GP smooth.

30 It captures if the chief executive is unitary (v2exhoshog=1) and directly elected by the population (v2expathhs=7) (Teorell and Lindberg,2015).

31 As included in Coppedge et al. (2020)

32 This measure was created by dividing the total number of autocratization and democratization episodes by the total number of countries per year (Maerz, Edgell, et al.,2021).

Regression analysis: results

Figure 4 gives estimated conditional relationships between autocratization and the four compo- nents of anti-pluralism. Each estimate comes from a separate model that conditions on all the contextual covariates described above. The Y-axis shows the expected change in the probability of autocratization and the X-axis the values of the component. We report additional model speci- fications in the Appendix, all of which give substantively similar estimates unless noted otherwise here.

Figure 4: Modeled Probability of Autocratization in the Year after Election (Indicator-level). Partial effects under Gaussian identity-link GAMs with GP smooths (𝑁 = 771). Each model includes the plotted index and adjusts for the same set of contextual coviariates (see above). Upper rugs show observations with autocratization at𝑡 + 1, lower rugs observations without it. 2016 US Republicans highlighted with a longer orange tick. Shaded±1SE and ±2SE regions.

0.0 0.2 0.4 0.6 0.8 1.0

−0.2 0.0 0.2 0.4 0.6

Lack of commitment to democratic process

GOP 2016

0.0 0.2 0.4 0.6 0.8 1.0 Disrespect for fundamental

minority rights

0.0 0.2 0.4 0.6 0.8 1.0 Demonization of opponents

0.0 0.2 0.4 0.6 0.8 1.0 Acceptance of political

violence

Weak commitment to the democratic process is associated with a substantially greater mod- eled probability of autocratization. This relationship is substantively and statistically significant starting already at relatively minor deviations from the democratic baseline (“[t]he party leader- ship was fully committed to free and fair, multi-party elections, freedom of speech, media, assem- bly and association”). We find a similar relationship to autocratization for acceptance of political violence. This association weakens if the other three party attributes (populism, cultural dimen- sion and economic left-right) are included in the model specification (see Figure C.10 in the Ap- pendix). Disrespect for minority rights has a similar, but weaker relationship to autocratization.

However, this relationship disappears if the other three party attributes are included in the model specification (see Table C.10 in the Appendix).

What is the substantive implication of these findings? For the litmus test to work, what mat-

ters most are pairwise associations of each characteristic with autocratization. This is what we

clearly find for all four characteristics, whether we adjust for contextual covariates or not. In short,

the socio-economic and political context does not seem to alter the picture. However, when con-

sidering the four at the same time, it seems that lacking commitment to democratic process and

accepting political violence are the most telling ones (Figure C.9 in the Appendix). Also extreme

levels of demonization of opponents seems to associate with a higher probability of autocratiza- tion in a given context.

Another perspective is available by considering the capacity to predict autocratization. We quantify it by Area Under the receiver operator Curve (AUC), which has a straightforward inter- pretation as the proportion of all possible {𝑦

= 0 , 𝑦

= 1 } observation pairs in which the latter has a larger predicted value of ( 𝑝

< 𝑝

). We estimate AUC with leave-pair-out cross valida- tion (Airola et al., 2009), by sampling 10 thousand {𝑦

= 0 , 𝑦

= 1 } observation pairs. A detailed summary features in Table C.1 in the Appendix. Lack of commitment to democratic process and acceptance of political violence already on their own achieve a fair predictive performance, with AUCs of 0.77, and adding all contextual covariates increases the AUCs only somewhat, to 0.83.

Figure 5: Modeled Probability of Autocratization in the Year after Election (Index-level) Partial effects under four Gaussian identity-link GAMs (𝑁 = 771). Each model includes the plotted index and adjusts for the same set of coviariates. Upper rugs show observations with autocratization at𝑡+1, lower rugs observations without it. 2016 US Republicans highlighted with a longer orange tick. Shaded±1SE and ±2SE regions.

0.0 0.2 0.4 0.6 0.8 1.0

−0.2 0.0 0.2 0.4 0.6

Anti−Pluralism

GOP 2016

0.0 0.2 0.4 0.6 0.8 1.0 Populism

0.0 0.2 0.4 0.6 0.8 1.0 Cultural Dimension

0.0 0.2 0.4 0.6 0.8 1.0 Economic Left−Right

Figure 5 shows the estimated relationships with autocratization of the Anti-Pluralism Index (API) and additional party characteristics under models that include socio-economic and political covariates. A greater level of anti-pluralism is associated with a greater modeled probability of autocratization. This relationship is substantially and statistically significant at high levels of anti- pluralism. For populist attributes some relationship to autocratization can be found; but only for very high levels of populism, for which we do not have many observations in the sample. For culturally far-right parties we find a positive relationship to autocratization onset as expected. In terms of economic party positions only extreme far-left victorious parties are associated with a statistically significant greater probability of autocratization.

Predictive model performance as captured by AUC-LPOCV, reported in detail in Table C.1 in

the Appendix, is better for the anti-pluralism either with (0.83) or without (0.79) the contextual

covariates. Populism, the cultural dimension, and economic left-right score is at 0.66, 0.61, and

0.62 without and at 0.82, 0.81, and 0.83 with the covariates. In short, the API on its own predicts

autocratization nearly as well as any of the other indexes combined with contextual covariates.